Self-driving cars have a hard time predicting bicycle movement, and workarounds that require cyclists to buy transmitters are running into resistance from some.
Via IEEE Spectrum:
"Deep learning is typically used for just detecting pixel patterns. We figured out an effective way to use the same techniques to estimate geometrical quantities," explains Deep3DBox contributor Jana Košecká, a computer scientist at George Mason University in Fairfax, Virginia.
However, when it comes to spotting and orienting bikes and bicyclists, performance drops significantly. Deep3DBox is among the best, yet it spots only 74 percent of bikes in the benchmarking test. And though it can orient over 88 percent of the cars in the test images, it scores just 59 percent for the bikes.
Slate reviewed the current workaround options:
One solution presented by Ford, Tome Software, and Trek Bicycle at CES last month is a concept known as bicycle-to-vehicle communications. Instead of just autonomous vehicles (or all motorized vehicles) on the road being able to wirelessly communicate their position and intentions with one another, bikes would be able to join the party. The proposed technology would be brand agnostic, something any cyclist could affix to herself or her bike. The key safety aspect of this connectivity would be that drivers would be alerted when a cyclist is nearby. It's similar, although potentially a step above, a concept presented by Volvo in 2014 that would work through tech embedded in a rider's helmet. Tome plans to hone its software, which could then be licensed out to vehicles, apps, bike accessories, and car accessories, at the Mcity autonomous driving test facility at the University of Michigan over the next year.
• The Self-Driving Car's Bicycle Problem (IEEE Spectrum via Slate)